François Brischoux, School of Biological Sciences A08, The University of Sydney, Sydney, NSW 2006, Australia. Tel.: +02 9351 3996; fax: +02 9351 5609; e-mail: firstname.lastname@example.org
Pupil shape in vertebrates ranges from circular to vertical, with multiple phylogenetic shifts in this trait. Our analyses challenge the widely held view that the vertical pupil evolved as an adaptation to enhance night vision. On functional grounds, a variable-aperture vertical pupil (i) allows a nocturnal species to have a sensitive retina for night vision but avoid dazzle by day by adjusting pupil closure, and (ii) increases visual acuity by day, because a narrow vertical pupil can project a sharper image onto the retina in the horizontal plane. Detection of horizontal movement may be critical for predators that wait in ambush for moving prey, suggesting that foraging mode (ambush predation) as well as polyphasic activity may favour the evolution of vertical pupil shape. Camouflage (disruption of the circular outline of the eye) also may be beneficial for ambush predators. A comparative analysis in snakes reveals significant functional links between pupil shape and foraging mode, as well as between pupil shape and diel timing of activity. Similar associations between ambush predation and vertically slit pupils occur in lizards and mammals also, suggesting that foraging mode has exerted major selective forces on visual systems in vertebrates.
Contrary to widely held opinion, the vertically slit pupil is not an adaptation to facilitate night vision (Walls, 1942; Detwiler, 1955). The major evidence contrary to this idea is the fact that under low light intensity, a dilated vertically slit pupil becomes round (Walls, 1942; Detwiler, 1955; see Werner, 1967 for pictures). Instead, the most crucial attributes allowing effective vision under dim light involve the ultrastructure of the eye, not pupil morphology. Strictly nocturnal animals possess retinas composed mainly of rods (black-and-white vision at low light intensities [scotopic vision, Crawford, 1934]), whereas the retina of diurnal species is composed primarily of cones (colour vision at high light intensities [photopic vision, Crawford, 1934]). Other characteristics of the eyes of nocturnal vertebrates include large and very curved corneas, large anterior chamber, large and spherical lenses (Detwiler, 1955), a reflective tapetum (Crawford, 1934), and pupils that are large relative to the focal lengths of the optical systems (Malmström & Kröger, 2006). Indeed, although the eyes of many nocturnal animals have vertical pupils, most strictly nocturnal species display round pupils (Walls, 1942).
Rather than enhancing visual acuity at low light levels, vertically slit pupils convey flexibility over aperture size (via facultative dilation) and hence increase the range of light intensities over which the eye can function effectively (especially important for species that are active both by day and by night: Walls, 1942). Closure of a round pupil (by a ring-shaped muscle) is less complete than closure of a slit pupil, which uses two additional muscles that laterally compress the pupil (Detwiler, 1955). Hence, a species with a round pupil adapted to dim light conditions may be unable to close its pupil enough to prevent dazzle under daylight conditions, whereas a vertically slit pupil may allow a nocturnal species to see well in daylight as well as at night.
The adaptive significance of the diversity of pupil shapes seen in vertebrates has attracted extensive speculation (Walls, 1942; Underwood, 1951; Heath et al., 1969; Hermann et al., 1975; Henderson & Binder, 1980; Zusi & Bridge, 1981; Murphy & Howland, 1986, 1990; Malmström & Kröger, 2006; Fontenot, 2008). Nonetheless, there appears to have been no analysis of ecological correlates of vertically slit pupils. Importantly, a vertically slit pupil might enhance visual acuity (Malmström & Kröger, 2006) as well as increasing the range of light intensities over which the eye can function effectively. First, a vertically slit pupil helps to maintain the length of focus and reduces blurring of the image. Animal eyes used under dim-light conditions are usually characterized by short depth of focus (Malmström & Kröger, 2006). As a consequence, possible chromatic defocus may blur the images (Kröger et al., 1999; Malmström & Kröger, 2006). Some vertebrates display multifocal lenses with concentric areas that focus different spectral ranges onto the retina, thereby solving the chromatic defocus problem (Malmström & Kröger, 2006). Aperture variation of a circular pupil will mask the marginal zones of the lens, cancelling the advantages of multifocal lenses. In contrast, slit pupils allow for the use of the complete diameter of the lens, even under high light intensities (Malmström & Kröger, 2006). Second, under daytime levels of illumination, light passing through a vertical pupil will exhibit different optical properties in the vertical versus horizontal plane. Because of the difference in aperture, the image will display a greater depth of field and a better sharpness in the horizontal than vertical plane (Heath et al., 1969). We suggest that these advantages are likely to be especially important for animals that lie in wait to ambush mobile prey. For most sit-and-wait foragers, prey will approach in the horizontal plane relative to the head of the predator. Vertical pupils might enhance fitness in ambush foragers for another, unrelated reason also: these predators rely on camouflage against the background to allow close approach by wary prey, and circular eyes often may be the most readily detectable body part (Walls, 1942). Vertical pupils may more effectively break up the outline of the head, rendering the predator less conspicuous.
In this paper, we test two hypotheses about the adaptive significance of pupil shape: that vertical pupils are adaptive to time of activity and to foraging mode. More specifically, we expect vertical pupils in nocturnal animals (because most need to be active at least occasionally by day, for activities such as thermoregulation or predator evasion) and in ambush foragers (because visual acuity in the horizontal plane is critical for foraging success in such taxa; and because vertical pupils may enhance camouflage). We use snakes as the lineage with which to test these predictions, because snakes share a similar ocular bauplan (Walls, 1942) because of their relatively recent evolution (apparently from burrowing forms with reduced eyes [Walls, 1940]; but see Caprette et al., 2004 for an alternative scenario involving aquatic ancestry), and pupil shapes, activity times and foraging modes vary extensively among species.
Materials and methods
We focused our analyses on Australian snake species belonging to the families Acrochordidae, Colubridae, Elapidae, Homalopsidae and Pythonidae. These groups were chosen because of the availability of published literature on activity and foraging, and access to preserved specimens at the museums. We excluded scolecophidians (blind snakes) because they are fossorial species with rudimentary eyes and presumably, very poor vision. To increase sample sizes, we also included in our dataset some related Elapid species from extralimital areas (Asian and other Pacific islands; see Table S1).
We scored pupil shape as circular, subcircular or vertically elliptical based on published literature (Boulenger, 1893, 1894, 1896; McDowell, 1969, 1970), and from photographs in books, field guides, and online (using the Google Image search tool http://www.images.google.com) and our own personal photographs (N = 79 species). For ambiguous cases, we examined preserved specimens at the Australian Museum (N = 48 species). Although characterization of pupil shape was straightforward on preserved specimens, we examined a mean of 6.6 specimens per species (± 2.6 SD) in case preservation had altered the dilation state of the pupil (Myers, 1984). In all cases, every specimen within a given species led to the same description of the pupil shape. In a few species (N = 11) examined using both methods (photographs and preserved specimens), both methods led to the same description of the pupil shape. Overall, scoring pupil shape as circular, sub-circular or vertically slit was straightforward (see Table S1).
Our final database consisted of a total of 127 snake species (including 7 non-Australian Elapids, see Table S1) for which we had information on pupil shape, 108 with data for activity times and 79 with data for foraging mode.
We ran two analyses: one to look at overall patterns among variables (i.e. treating species as independent units) and one to look at functional links (i.e. to examine whether phylogenetic changes in one trait were nonrandomly associated with concurrent shifts in other traits). First, we used nominal logistic regression to examine the relationship of pupil shape (independent variable) with foraging mode and activity (dependent variables), with Family as a covariate. We ran these analyses in JMP 7, using our complete dataset of 127 snake species. Second, we used phylogenetic comparative correlation to analyse our data. We based our phylogenetic tree on published hypotheses (Kelly et al., 2003: Caenophidia; Keogh et al., 2003: Hoplocephalus spp.; Scanlon & Lee, 2004: Elapidae; Wüster et al., 2005: Pseudechis spp.; Vidal et al., 2007: Caenophidia; Alfaro et al., 2008: Homalopsidae; Rawlings et al., 2008: Pythonidae; Sanders et al., 2008: Australasian Elapidae).
Tropidonophis mairii (Natricinae) and Stegonotus spp. (Colubrinae) were not included in any of the published phylogenies. We placed T. mairii as sister of the Colubrinae because this taxon was the only Natricine in our database (relationships between the two sub-families from Kelly et al., 2003 and Vidal et al., 2007). Some Stegonotus spp. (including S. cucullatus) were previously classified as Lycodon (see Boulenger, 1893). Because of the close morphological similarities of these two genera (compared to the other Colubrinae included in our study), we considered them as sister groups in our analyses.
Because the published phylogenetic hypotheses we used to construct our tree had only minor overlap, there were very few conflicts, and it was straightforward to combine them. All the hypotheses used (except for that of Scanlon & Lee, 2004) were based on analyses of molecular data, usually the same genes; hence, the hypotheses were not conflicting when the same taxa were used in more than one paper. The only conflicts involved taxa that were not included in our analyses (because data on relevant variables were lacking). Finally, the groups involved in our analyses all have well-resolved phylogenies (excepted for the Colubridae, which represent only 5% of the species in our study).
We reconstructed the ancestral states of pupil shape, activity rhythm and foraging mode for 89 snake species (i.e. those that were included in a phylogenetic hypothesis). To check that our conclusions were not unduly affected by omitting non-Australian snake taxa within some of the lineages we studied, we repeated the comparative analyses using the most intensively sampled lineage (Elapid snakes) only. Both analyses resulted in similar trends (see Results). However, because Elapids constitute a more homogeneous group, phylogenetic analyses for this group were more robust. As a consequence, we have emphasized the results concerning Elapids (see Results).
We reconstructed ancestral states using parsimony and treated the states of all characters as ordered (sub-circular pupils, mixed foraging mode and polyphasic activity were considered as transitional states). Relationships between pupil shape and diel time of activity, and between pupil shape and foraging mode, were investigated using Pagel’s correlation analysis for nominal characters (Pagel, 1994). Because this method requires binary characters (Pagel, 1994), we ran the analyses using all possible combinations of groupings (i.e. intermediate states plus terminal states). For each correlation, we used eight parameter models, with an intensity of likelihood search equal to ten and 1000 simulations (as suggested by Midford & Maddison, 2009). All phylogenetic correlations and ancestral state reconstructions were estimated using Mesquite 2.6 (Maddison & Maddison, 2009).
The logistic regression on 127 snake species identified a significant relationship between pupil shape and foraging mode (likelihood ratio χ2 = 44.10, df = 15, P < 0.0001), and between pupil shape and diel time of activity (χ2 = 96.07, df = 24, P < 0.0001). Families differed in the relationship between pupil shape and foraging mode (χ2 = 39.90, df = 25, P = 0.03) and between pupil shape and time of activity (χ2 = 67.05, df = 30, P = 0.0001). The same patterns were seen when analyses were based on Elapids only (n = 85 species, incorporating substantial variation in all characters: foraging mode: χ2 = 34.94, df = 10, P < 00.0001; activity: χ2 = 91.51, df = 24, P < 0.0001).
Most species with vertical pupils were nocturnal ambush foragers (Fig. 1, group A). Snakes with circular pupils were mainly active foragers, and most species were diurnal (Fig. 1, group D) although many actively foraging snakes with circular pupils were nocturnal or polyphasic (Fig. 1, groups B and C). The other well-represented groups were active-foraging species with sub-circular pupils and nocturnal activity, and ambush-foraging snakes with vertical pupils and polyphasic activity cycles (Fig. 1, groups E and F).
The ancestral Elapids were probably diurnal active foragers with circular pupils (Fig. 2). These traits also are ancestral to the Australian Elapids, but with pupils possibly sub-circular rather than circular (Fig. 2). The evolutionary history of pupil shape and activity times remains unresolved in some clades of Australian Elapids (Fig. 2). Vertical pupils evolved twice and sub-circular pupils at least twice (and up to six times) within this group. Polyphasic activity evolved at least five times, with at least four reversions to diurnality (Fig. 2). Ambush foraging evolved independently four times (in Acanthophis, Denisonia, Echiophis and Hoplocephalus), but most Elapids retain the ancestral state of active foraging.
If phylogenetic relatedness is factored out of the analysis using Pagel’s correlation methods (on the entire dataset), pupil shapes were strongly linked to foraging mode (P < 0.005 in all cases) and activity times (P < 0.0001 in all cases) regardless of how we grouped species with intermediate states of foraging mode and activity times. Considering Elapids only, pupil shape was always functionally linked to foraging modes (P < 0.018) and activity times (P < 0.034), except when nocturnal activity was pooled with polyphasic (P = 0.11 for vertical pupil grouped with sub-circular, and P = 0.11 for sub-circular pupil grouped with round pupil).
Among Australian snakes, pupil shapes correlate both with diel activity times and with foraging modes. Ambiguity remains about ancestral states of pupil shape for the Elapid family, but it is clear that pupil shape has changed repeatedly among the Australian Elapids. As we predicted (see Introduction), most snake species with vertical pupils are nocturnal and also are ambush foragers, and most snakes with circular pupils are diurnal and active foragers. Overall, foraging mode predicts pupil shape accurately in more cases than does diel time of activity, because many active-foraging snakes with circular pupils are not diurnal.
Although our analysis supports the idea that foraging mode and activity times both impose selective forces on pupil shape in snakes, the actual selective forces remain unclear. The relationship to time of activity is likely to reflect the ability of species with vertical pupils to see more clearly over a wide range of light intensities (Walls, 1942; Detwiler, 1955). Even a “nocturnal” species likely benefits from being able to see prey and predators by day. For an exclusively diurnal species, however, the advantages of colour vision at high light intensities may favour optical systems that function well only in daylight hours.
The link between vertical pupils and foraging mode is more difficult to understand in causal terms. We suggested two reasons why vertical pupils would enhance fitness in ambush foragers: enhanced visual acuity for movements on a horizontal plane and/or camouflage through disrupting the normal circular shape of the eye. The vertically slit pupil allows the projection of a sharper and deeper image onto the retina in the horizontal plane because of the differences in aperture (Heath et al., 1969) and also maintains the length of focus and reduces blurring of the image (Kröger et al., 1999; Malmström & Kröger, 2006). These visual advantages apply by day and at dusk and dawn, but not late at night when the dilated state of the pupil should result in acuity similar to that of a round pupil (Walls, 1942; Detwiler, 1955; see Werner, 1967 for pictures). Our second (camouflage) hypothesis concerning an advantage of vertical pupils for ambush foragers is consistent with (rather than an alternative to) the visual-acuity hypothesis. Further work could usefully examine the detectability of snake models with vertical versus circular pupils.
As noted earlier, a vertical pupil allows a species to see more clearly over a wide range of light intensities. Therefore, one should expect species with vertical pupils to be polyphasic rather than strictly nocturnal. Why, then, do our results suggest a link between nocturnality and vertical pupils? First, a nocturnal species might need to be active at least occasionally by day, for activities such as thermoregulation or predator evasion. Second, published categorizations of activity times in snakes rely on a complex interaction between activity, behaviour, thermal conditions and detectability of these secretive animals; and this is especially true for cryptic ambush foragers. It may be difficult for a human observer to assess the “activity” of an immobile, concealed snake that lies in wait to ambush mobile prey (either during the day or at night). Additionally, we suspect that the widely held (but incorrect) opinion that the vertically slit pupil is an adaptation to facilitate night vision has affected the accuracy of published reports of diel activity times in many species. That is, some authors may have inferred nocturnal activity based on the vertical pupil shape, rendering these data useless for our analysis. Because of their distinctive foraging strategy, we suggest that many ambush-foraging snakes are likely to be “active” (i.e. ready to seize prey) both during the day and at night. Such taxa should be categorized as polyphasic or “arrhythmic” rather than as strictly diurnal or nocturnal.
A more detailed examination of pupil shapes and foraging modes in snakes provides additional insights into the link between those traits. For example, three major (familial-level) snake lineages (Boidae, Pythonidae, Viperidae) mostly contain heavy-bodied species that rely on ambush predation; almost all have vertically slit pupils (Greene, 1997). The only exception is among Viperids, wherein nocturnal species from one phylogenetically basal genus Causus (“night adders”) have circular pupils. Interestingly, these species also forage actively for anuran prey (Ineich et al., 2006). Viperids, Boids and Pythonids also are the only snake families to have evolved an additional sensory mode: facial pits that detect small thermal differentials (Greene, 1997) used to locate warm-blooded prey (Shine & Sun, 2003). That convergence suggests strong selection for an ability to detect and localize prey in ambush-foraging snakes, and hints that visual acuity may be under similarly strong selection in such taxa (Greene, 1997).
Such functional advantages of a vertical pupil for ambush foragers are unlikely to apply only to snakes. Examples from other lineages suggest that the link between pupil shape and foraging mode may be widespread. For example, Pygopodid lizards are highly modified legless gekkonoids with remarkable morphological and ecological convergences with snakes (Wall & Shine, 2007). Interestingly, the only genus of this family to rely on ambush foraging (Lialis spp.) is also the only one characterized by vertically slit pupils (Reilly et al., 2007; Wall & Shine, 2007). Divergence in pupil shapes among carnivorous mammals (Malmström & Kröger, 2006) may accord with foraging mode also. Among Felids, domestic cats are ambush foragers with vertical pupils (Murray et al., 1995), whereas several larger Felid species have round pupils and are active foragers (Sunquist & Sunquist, 1989). Among Canids, the Red Fox is an ambush forager with vertically slit pupils, whereas the Grey Wolf is an active forager with round pupils (Henry, 1986). Future research could usefully extend our analyses to a broader range of vertebrate lineages. The broad similarities in visual-perception challenges faced by ambush foragers, and in the basic structures of the vertebrate eye, create exciting opportunities to test general hypotheses about the role of foraging mode in exerting selection on specific features of eye anatomy in vertebrates.
We thank Sylvain Dubey for discussions about vertical pupils in snakes and lizards, Greg Brown for suggesting the camouflage hypothesis, Hebert Ferrarezzi for suggestions and discussion on the reconstruction of ancestral states and Marine Danek-Gontard for comments on an earlier draft of the manuscript. We thank Ross Sadlier and Cecilie Beatson for permission to examine specimens under their care at the Australian Museum (Sydney). Funding was provided by the Australian Research Council (ARC) and the Australian Government (Endeavour Award # 930_2009).